The setup
Lumen & Loam is one of three internal Hookframe portfolio brands — built end-to-end as proof that the creative engine works on real spend, in a real ad account, on a real Shopify store. No client involved. No outside budget. Just us, eating our own dog food.
Hero SKU: a $34 men's daily face wash. Brand vibe: minimalist sandstone palette, derm-credible photography, no testosterone-bro tropes. Target: men 25–44, DTC-curious but skincare-skeptical.
The hypothesis
Most men's skincare ads fail because they assume men want to be told they're broken. Our hypothesis: lead with the identity tension — "you're already doing skincare, you just don't know it" — instead of the deficit ("you have wrinkles, fix them").
What we shipped
- Six angle buckets: identity-tension, dad-routine, dermatologist authority, before-after, contrast vs. women's brands, time-savings ("30 seconds")
- 28 finished variants across 9:16 + 1:1 + 16:9
- Generation stack: Higgsfield-class for cinematic intro frames, Veo for product realism, Runway Gen-4 for motion fills
- Captioning: brand-styled, burned in, white-on-near-black with sandstone accent
- Meta-ready filenames matching the test ABO/CBO structure
What worked
Three hooks beat the incumbent product-demo cut by Day 14:
- "My dermatologist quit." — Identity-tension narrative. CTR 2.8%, CPA $26.
- "$300 routine in one bottle." — Stat-led contrast vs. typical women's-brand bundles. CTR 2.4%, CPA $29.
- "30 seconds. Once a day." — Time-savings demo with cinematic foam macro. CTR 2.1%, CPA $31.
The dermatologist-authority angle (the conventional choice) underperformed at CPA $51 — a useful kill.
What didn't work
- "Cinematic close-up product unboxing" — beautiful, useless. CTR 0.7%, no purchases.
- "Dad-bod glow-up" — too on-the-nose. Felt like a parody, didn't resonate.
- 16:9 cuts on Meta were a waste of inventory. 9:16 + 1:1 carried the whole test.
The lesson
Volume of testing wins. None of us would have predicted "my dermatologist quit" as the runaway hook from a moodboard. The whole point is you don't have to predict — you ship enough variants that the algorithm tells you.
What we'd do differently
- Skip 16:9 next time, reallocate that production budget into 5 more 9:16 variants
- Test a UGC-style AI talking-head treatment as an angle bucket (we didn't this round)
- Front-load the "30 seconds" mechanic into the first 0.4s — it currently shows at 1.1s, which is too late
All figures are from our internal Meta + TripleWhale-lite alternative dashboards. Lumen & Loam is a Hookframe-owned portfolio brand. Real store, real spend, real receipts — clearly labeled.